CINECA Guidelines for the submission of proposals
This set of guidelines will help you in the submission of your application to receive an award of HPC resources at CINECA. All the proposals, once submitted, will be evaluated within CINECA Peer Review system. The proposals must be submitted using the web available at https://userdb.hpc.cineca.it.
The following text describes the general structure of the proposal that must be completed to apply for an ISCRA project. Notice that sections and subsections numbers in this document correspond to items with the same references in the web form (e.g. subsection 4.1 here, corresponds to text field 4.1 in the form)
The web form may vary slightly from the describe one. Class C template is simplified.
1. Personal data of PI and Research Group
This section explains which personal information are requested to submit a project (“Section 1” of the submission web form). It is requested for all projects classes (B, C).
The Principal Investigator (PI) is the only responsible for the project and for the management of any associated resource. If your project has multiple investigators, you can list all of them in the Resaerch Group section. The PI must be working in an Italian University or in an Italian Research Agency. In any case the work must be performed at an Italian institution.
2. Your Experience
2.1 List of Publications
At least the four most relevant publications produced over the last 5 years by the PI or the group must be listed.
2.2 Curriculum Vitae
The proponent is expected to provide a Curriculum Vitae (max 4000 chars) for the PI and any relevant Co-PI listed on the application.
2.3 Outcomes of previous approved projects at CINECA
The proponent MUST describe the main achievements reached during any previous grant received from CINECA. Describe the PI previous activities with regard to any projects carried out in HPC environment at CINECA.
Descriptions of the PI previous activities, particularly with regard to his/her experience in supercomputing and any projects carried out in HPC environments (e.g. with CINECA, or using infrastructures such as those of DEISA or PRACE). The applicants should declare if they have submitted similar projects in other contexts (e.g. DECI or other computer centres) and, if so, say which ones
3. Institutional Contact
The Institutional Contact mentioned here, is a person with a permanent positon.
This section explains how to write the abstract/summary of the project (“Section 3” of the submission web form). It is requested for all projects classes (B, C).
o You are requested to specify the project title and acronym (8 characters max.).
o Select, among those available in the web-menu, the science category which your project belongs to.
o Provide an executive summary that accurately describes your proposed research, focusing on the following topics:
• project objectives;
• scientific rationale;
• innovation potential;
• state of the art;
• outcomes and high-impact scientific advances expected.
This section cannot exceed 4000 char.
Only for C class projects, please describe here the reasons for requesting specialist support for the enabling and optimization of applications.
5. Research Objectives
This section explains how to detail the project, both from a scientific and from a technical point of view (“Section 4” of the submission web form). It is requested only for projects classes B.
5.1 Description of Research
This section, including references, cannot exceed 20.000 char and is expected to detail how the specific scientific/computational goals will be achieved and to define detailed workplan.
Proposals will be evaluated on both scientific and technical merit. The provided information should be sufficient for the reviewers in your research field to provide a scientific evaluation of the proposal and to understand if the computational methodology is suitable to reach the project’s goals. Furthermore, a general scientific cross-comparison with proposals in other disciplines should be feasible.
The list of the topics that MUST be detailed/included follows (please notice that incomplete descriptions will lead to the project rejection).
• Scientific framework
• Project objectives
• List of milestones
• Theoretical and computational methods employed
• List of the applications to be used and their performance on parallel architectures (scalability and load-balancing)
• Related Publications
• Detailed workplan and timetable of the activities (GANTT)
• Any relevant figure and image
5.2 Significance of Research
Place the proposed research in the context of competing work in your discipline. In addition explain what scientific advances you expect to be enabled by an award that justifies an allocation of large-scale resources. This section, including any references you provide should not exceed 8000 char.
6. Computational Approach
This section explains how to detail the methodological approach to the project, (“Section 5” of the submission web form). It is requested only for projects classes B.
6.1 Computational approach.
Provide quantitative evidence of the HPC performances of the production application you will adopt in the project (scalability, efficiency, I/O performances). Parallel performances in either strong or weak scaling mode should be provided. Weak scaling behaviours are probed by holding per-processor computational work constant (e.g., the size of the mesh on a processor is held constant) as the total problem size grows with number of processors. Strong scaling behaviours are probed by holding the total problem size constant as the processor count grows, thereby decreasing the per-processor computational work. Benchmark data should be provided in either tabular or graphical form, or both; the speedup curve should be supplied as well for strong scaling examples. Where appropriate, characterize the application’s single-node performance (ex. percent of peak).
This section, including any references you provide cannot exceed 16.000 chars.
6.2 Application Packages
List each application or software package that will be used in the project, including post-processing packages. For each package, specify if it is open, proprietary or licensed and the communication and library requirements.
6.3 Specialistic support
Class B proposals: Please describe here the reasons for requesting specialist support for the enabling and optimization of applications. The request will be evaluated by CINECA.
Class C proposals: Please describe here the reasons for requesting specialist support. The request will be evaluated by CINECA.
7. Requested Resources
In this section, you are expected to give details on the following items.
It is requested only for projects classes B.
7.1 Optimal number of core hours
• Computer architecture requested
• Minimum and optimal number of CPU hours in order to complete the project
Please fill in the form considering the samples drafted here for the calculation of minimum and optimal number of CPU.
Minimum number of core hours = (elapsed time of a single run)* (# of cores used in a single run) * (total # of runs).
Minimum number of core hours: 1000*128*12=1,536,000 CPU hours
The optimal number represents the CPU time necessary to reach all the targets of the project.
The minimal number is the CPU time necessary to accomplish a minimum meaningful work.
7.2 Requirements for a typical run
• Maximum number of usable cores
• Minimal memory per core (GB) at maximum number of cores
• Total memory for smallest target problem (GB)
• Total memory for largest target problem (GB)
7.3 I/O Requirements:
o Restart I/O: Describe I/O requirements for program restart. Include a list of the types of files you write out for program restarts, i.e. MPI/IO, raw binary, serial from all processors, HDF5, etc. and the following quantitative information: amount of restart data written out, or percentage of memory, how many processors write-out restart data, how much memory your job takes per processor, the size of a single restart dump, maximum percentage of total time acceptable time to read/write restart data.
o Analysis I/O: Describe your I/O requirements for analysis. Include a list of the types of files you write out for analysis, i.e. HDF5, NetCDF, PHDF5, PnedtCDF; size of analysis dump, amount and type of data read into analysis program, etc.
o Archival I/O: Describe your archival data requirements. Include details on the following: total space needed; file size distribution (many small files or one large file), access pattern, retention needs and frequency of access.